Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on MiniMax M2.5/M2.7. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Near the low end of the 31–132 tok/s/user interactivity band, at 56 tok/s/user on MiniMax M2.5/M2.7: B300 runs 7774 tok/s/chip at $0.08/M tokens, GB200 NVL72 runs 7412 at $0.07/M. GB200 NVL72 is 16% cheaper per token; B300 delivers 5% more tok/s/chip.
Setting 81 tok/s/user as the target on MiniMax M2.5/M2.7, B300 produces 4243 tok/s/chip ($0.15 per million tokens) and GB200 NVL72 produces 3276 ($0.16). B300 is 7% cheaper per token; B300 delivers 30% more tok/s/chip.
At 107 tok/s/user interactivity on MiniMax M2.5/M2.7, B300 delivers 2226 tok/s/chip at $0.28 per million tokens; GB200 NVL72 delivers 1964 tok/s/chip at $0.26. GB200 NVL72 is 7% cheaper per token; B300 delivers 13% more tok/s/chip at this point. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:7774.5GB200 NVL72:7412.3 | B300:4242.7GB200 NVL72:3276.0 | B300:2226.1GB200 NVL72:1964.3 |
| Cost ($/M tok) | B300:$0.081GB200 NVL72:$0.070 | B300:$0.148GB200 NVL72:$0.158 | B300:$0.282GB200 NVL72:$0.263 |
| tok/s/MW | B300:4091823GB200 NVL72:3963793 | B300:2233015GB200 NVL72:1751893 | B300:1171652GB200 NVL72:1050451 |
| Concurrency | B300:~335GB200 NVL72:~92 | B300:~6GB200 NVL72:~30 | B300:~13GB200 NVL72:~13 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr: VR200: 3.61H100: 1.17H200: 1.22B200: 1.73B300: 2.26GB200: 1.86GB300: 2.31MI300X: 0.95MI325X: 1.1MI355X: 1.5RTX6000PRO: 0.68Jalapeño (Teacup): 1.47
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
